David Hartley, Director, Insurance Fraud, SAS Global Security Intelligence Practice
The insurance industry’s transition to online business processes has been welcomed by customers. Researching rates and submitting policy applications has never been so easy. However, fraudsters use this to their advantage. The introduction of digital application processes has made it easier to commit insurance application fraud in three surprisingly simple ways:
- Fraudsters open insurance policies for fictitious beneficiaries
- Agents skim premiums or open and cancel policies to make quota and bonuses
- Customers tweak and falsify application information to reduce their premiums
However, while digital channels open new opportunities for insurance application fraud, they also provide new data points to fight back. Smart insurers will take advantage of data from device fingerprints, IP addresses, geolocation to help detect fraud as early as the point of application. There are various effective defence strategies powered by analytics and machine learning.
Authenticate identity digital application stage
Insurers have masses of data, both internally and externally that can help determine the legitimacy of an application and the individual behind it. There are effective, data-driven approaches to combat insurance application fraud:
- Monitor application data to see if the same information or device is being reused across multiple identities that otherwise look unrelated
- Assess past experience for existing or cancelled policies that share a data element with the new application, such as the same device ID, address or SSN
- Find “proof of life,” details you would associate with a real personal identity, such as driver’s licence, voter registration or property ownership
- Analyse the network surrounding the application, looking for unusual or suspicious connections (or lack thereof) among applicants, devices, policies and application data
Avert potential claims fraud at the point of application
What if you could stop claims fraud before it ever had a chance to get started? What if you could use intelligence gained from the claims detection process to better understand new applications and tag the right ones to send directly to investigative teams?
You can. Network analytics connects the dots by spotting links between prior claims alerts and new applications. What did applications that led to fraud look like in the past, and does this application look similar? Connections can be established not just through people or vehicles (addresses, phone numbers, VINs, etc.), but through any number of attributes, such as IP addresses, devices, bank accounts, repair shops and medical providers.
You can enrich the discovery by applying key scenarios learned from prior claims processing. Through machine learning, a form of artificial intelligence, new information gained through analysis can be fed back into models for continuous improvement.
Recognise agent gaming
Unscrupulous agents have a variety of options for gaming the system for personal benefit. For example, they can take premiums from customers without filing the policies with carriers or persuade customers to buy unnecessary coverages to earn extra commissions.
A well-rounded fraud solution uses multiple analytic techniques to lead you to agents who might warrant a closer look. Industry-leading solutions detect agent gaming using the following techniques:
- Machine learning examines patterns of agent behaviour and matches it with scenarios that past experience shows were associated with gaming
- Peer grouping clusters agents with similar attributes – career level, specialisation, region to better compare their activity to that of their peers
- Anomaly detection finds agents who are performing quite differently from their peers or who show a radical shift in activity that may point to agent gaming
- Social network analysis connects the dots around an agent, revealing links and overlaps among entities in the application, such as households, VINs and insured properties
Together, these techniques can highlight unusual patterns while reducing the chance of false positives and unnecessary inquiries. Ultimately the system learns from each experience and its outcomes to continuously improve the analytics each time you run the cycle.
Pinpoint customer gaming at the point of application
Insurers that sell primarily over the telephone or internet are subject to a number of well-known and emerging threats. Customers can falsify information about the primary driver or where the vehicle will be garaged or acquire refunds on cancelled policies that were bought with stolen cards.
Analytics can spot this form of gaming in real time by setting thresholds that define how much an applicant can manipulate the premium before triggering action, such as messaging, call-backs or blocks. Analytics can also detect unusual patterns of activity, such as multiple policy cancellations linked to the same device.
Insurance application fraud analytics
Insurance companies that have invested in strong anti-fraud capabilities have seen dramatic results. For example, SAS partnered with a large US carrier to deploy a fraud solution to identify agent gaming and increase the productivity and throughput of field underwriting, territory managers and internal audit teams.
Using analytics to improve detection, the solution found ten times worse-performing agents – forty per cent referred to internal audit, compared to only four per cent under the legacy process. At the same time, analysis and investigation efficiency improved by thirteen hours, and data gathering efficiency improved by two hours.
The solution collates the data enabling businesses to complete five investigations in the time it used to take for one. Analytics provides actionable insights allowing audit teams to focus on fundamental work as opposed to data organisation.
Ultimately, having a robust application and claims fraud solution enables you to say goodbye to fraud before it’s too late.
